CN115545355B - Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition - Google Patents
Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition Download PDFInfo
- Publication number
- CN115545355B CN115545355B CN202211512825.2A CN202211512825A CN115545355B CN 115545355 B CN115545355 B CN 115545355B CN 202211512825 A CN202211512825 A CN 202211512825A CN 115545355 B CN115545355 B CN 115545355B
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- power grid
- characteristic data
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000004927 fusion Effects 0.000 title claims abstract description 32
- 238000003745 diagnosis Methods 0.000 title claims abstract description 29
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 86
- 238000012544 monitoring process Methods 0.000 claims description 33
- 230000015654 memory Effects 0.000 claims description 27
- 238000001454 recorded image Methods 0.000 claims description 25
- 238000000605 extraction Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 10
- 238000013527 convolutional neural network Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims 1
- 239000000758 substrate Substances 0.000 claims 1
- 239000010410 layer Substances 0.000 description 14
- 210000004027 cell Anatomy 0.000 description 10
- 230000008569 process Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 239000011248 coating agent Substances 0.000 description 5
- 238000000576 coating method Methods 0.000 description 5
- 239000000428 dust Substances 0.000 description 5
- 239000000779 smoke Substances 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000002457 bidirectional effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000002355 dual-layer Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
Abstract
The invention discloses a power grid fault diagnosis method, device and equipment based on multi-class information fusion identification, wherein the method comprises the following steps: preprocessing various types of data in a power grid to obtain characteristic data of the various types of data; fusing the characteristic data of the multiple types of data to form fused characteristic data; and analyzing the power grid equipment faults according to the fused characteristic data. The present disclosure contemplates that different types of grid data can provide a more accurate data basis for grid fault diagnosis.
Description
Technical Field
The invention belongs to the technical field of intelligent diagnosis of power grid faults, and particularly relates to a power grid fault diagnosis method, device and equipment based on multi-class information fusion identification.
Background
With the proposal of the policy of constructing a novel power system taking new energy as a main body, a power grid presents new characteristics of high voltage level and high proportion of new energy grid connection, so that the uncertain factors of power grid operation are enhanced, and unprecedented new requirements for power grid fault cause diagnosis and treatment are put forward. The current power grid operation information stores a large amount of protection wave-recording image data, monitoring operation information data and other data of various types, and the data can be used for quickly tracing the fault reasons of power grid equipment, so that a strategy generation basis is provided for power grid fault treatment and recovery.
However, in the prior art, due to the lack of effective fusion and analysis means for the various types of data, in the power grid fault detection, fault monitoring information and protection wave-recording image information are more accurate data for reflecting the power grid fault, but in the prior art, the two types of data are not fused to be considered for comprehensive analysis. When faults occur, a plurality of schedulers rely on experience to comprehensively analyze and judge fault causes, so that a large amount of valuable data is difficult to fully utilize, and meanwhile, the power grid scheduling is excessively dependent on manual experience, so that intelligent scheduling is difficult to realize. Meanwhile, due to the fact that a dispatcher needs to combine multiple types of data analysis and calculation artificially, the fault cause of the power grid equipment is difficult to diagnose accurately in time, reasonable prevention and control measures are given, and the safe and stable operation of the power grid is threatened under serious conditions.
Disclosure of Invention
In order to solve the problems, the invention provides a method, a device and equipment for carrying out fusion identification on various types of information so as to diagnose power grid faults.
The disclosure provides a power grid fault diagnosis method based on multi-class information fusion identification, which comprises the following steps:
preprocessing various types of data in a power grid to obtain characteristic data of the various types of data;
fusing the characteristic data of the multiple types of data to form fused characteristic data;
the fused characteristic data comprise mixed coding vectors and fault class marks of the multiple types of data;
and analyzing the power grid equipment faults according to the fused characteristic data.
In some embodiments, after the fused feature data is formed, performing logic operation on the fused feature data and the predicted feature data to obtain a fault cause prediction error at the previous moment;
determining an input gate signal, a forget gate signal and an output gate signal in a gate structure in an M-LSTM network based on the fault cause prediction error at the previous moment and the fault feature vector at the current moment;
and finally, determining the fault of the power grid equipment.
In some embodiments, the fault cause prediction error at the previous time is obtained by:
wherein ,
In some embodiments, the predictive value for the plurality of types of data is obtained by an LSTM algorithm.
In some embodiments, the ingress gating signals, the forget gating signals, and the egress gating signals in the gating structure in an M-LSTM network are determined by:
wherein ,
I t for inputting gate signals for controlling the input of fault signature sequences in memory cells in LSTM,F t for forgetting the gate signal, for determining how much information is kept in the memory cell at the previous time in the current LSTM,O t determining the output content of the LSTM as an output gate;
W i input gate weights in the LSTM algorithm; x is x t Input data at the time t; b i An input gate bias term;
is a sigmoid function; w (W) f Forgetting the door weight; b f A forget door bias term; w (W) o Outputting door weights; b o To output a gate bias term.
In some embodiments, the memory cells in the LSTM are updated by
：
wherein ,W c is implicit state weight;b c biasing items for implicit states.
In some embodiments, the mixed coding vector in the fused feature data is used as an input of M-BiLSTM;
outputting a fault reason category number by using the M-BiLSTM;
and determining the power grid fault reason according to the fault reason category number.
The disclosure also provides a power grid fault diagnosis device based on multi-class information fusion identification, the device comprising:
the device comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for preprocessing various types of data in a power grid and acquiring characteristic data of the various types of data;
the fusion unit is used for fusing the characteristic data of the multiple types of data to form fused characteristic data; the fused characteristic data comprise mixed coding vectors and fault class marks of the multiple types of data;
and the analysis unit is used for analyzing the power grid equipment faults according to the fused characteristic data.
In some embodiments, after the fused feature data is formed, the analysis unit performs a logic operation on the fused feature data and the predicted feature data to obtain a fault cause prediction error at a previous moment;
determining an input gate signal, a forget gate signal and an output gate signal in a gate structure in an M-LSTM network based on the fault cause prediction error at the previous moment and the fault feature vector at the current moment;
and finally, determining the fault of the power grid equipment.
In some embodiments, the fault cause prediction error at the previous time is obtained by:
wherein ,
In some embodiments, the predictive value for the plurality of types of data is obtained by an LSTM algorithm.
In some embodiments, the ingress gating signals, the forget gating signals, and the egress gating signals in the gating structure in an M-LSTM network are determined by:
wherein ,
I t for inputting gate signals for controlling inputs into memory cells in LSTMThe sequence of barrier features,F t for forgetting the gate signal, for determining how much information is kept in the memory cell at the previous time in the current LSTM,O t determining the output content of the LSTM as an output gate;
W i input gate weights in the LSTM algorithm; x is x t Input data at the time t; b i An input gate bias term;
is a sigmoid function; w (W) f Forgetting the door weight; b f A forget door bias term; w (W) o Outputting door weights; b o To output a gate bias term.
In some embodiments, the memory cells in the LSTM are updated by
：
wherein ,W c is implicit state weight;b c biasing items for implicit states.
In some embodiments, the mixed coding vector in the fused feature data is used as an input of M-BiLSTM;
outputting a fault reason category number by using the M-BiLSTM;
and determining the power grid fault reason according to the fault reason category number.
The present disclosure also provides an electronic device comprising at least one processor and a memory electrically connected to the processor, wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of the preceding claims.
Compared with the prior art, the invention has the following advantages:
because the multisource and different types of data such as the wave-recording image, the monitoring information and the like are fused, a more accurate data base is provided for subsequent comprehensive fault diagnosis;
by considering the prediction errors of the actual measured value and the predicted value, the fault diagnosis is carried out by combining the data base, so that the running state of the power grid when the power grid equipment fails is more comprehensively and deeply reflected;
by establishing the power grid equipment fault cause diagnosis model, the equipment fault cause diagnosis accuracy is improved, and the method has an important role in reducing the power grid cascading risk caused by faults and improving the quick power supply capacity of the power grid equipment.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 illustrates a general basic flow diagram of a method for diagnosing a power grid fault based on multi-class information fusion identification in accordance with an embodiment of the present disclosure;
FIG. 2 illustrates a basic flow chart of a grid fault diagnosis method for fusion identification of fault monitoring information and recorded image information in accordance with an embodiment of the present disclosure;
FIG. 3 shows a schematic diagram of a power grid fault diagnosis apparatus according to an embodiment of the disclosure;
fig. 4 shows a schematic structural diagram of an electronic device according to an embodiment of the disclosure.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a basic flowchart of a power grid fault diagnosis method based on multi-type information fusion recognition according to an embodiment of the present disclosure, as shown in fig. 1, in an embodiment of the present disclosure, after acquiring multiple types of data in a power grid, preprocessing the multiple types of data in the power grid to obtain characteristic data of the multiple types of data; then, fusing the characteristic data of the multiple types of data to form fused characteristic data; and finally, analyzing the power grid equipment faults according to the fused characteristic data.
In the embodiment of the present disclosure, a plurality of types of data (illustratively, a recorded image, monitoring information, etc.) including two or more types of data among image type data, picture type data, text type data, etc. are subjected to fusion processing. Compared with the prior art that only one type of data is considered, the data fusion method and device based on the data fusion system are capable of comprehensively and accurately identifying faults when different types of data are considered.
It should be noted that, in the embodiment of the present disclosure, the protection recorded image and the monitoring information are taken as two different types of data among a plurality of different types of data as an example for illustration, but the present disclosure is not limited to only these two types of data, and any data as long as it is two or more different types of data can be applied to the present disclosure.
In the embodiment of the disclosure, fault monitoring information and protection record image (also referred to as fault protection action record chart) information are obtained, where the fault monitoring information may include: one or more data of fault equipment description, fault time, power flow abrupt change equipment description, power grid operation mode information description, meteorological information and the like; the protective recording image information may include: event time, current amplitude phase, voltage amplitude phase, etc. It should be noted that, in the embodiments of the present disclosure, the data is not limited to the above, and other data capable of describing the fault-monitoring information and the protection recorded image information may be applied to the present disclosure.
1. In the embodiment of the disclosure, a multi-source data preprocessing module can be adopted to process fault monitoring information and protection recorded image information. Wherein the fault monitoring information includes: one or more of the factors including fault equipment description, fault time, power flow abrupt change equipment description, power grid operation mode information description, meteorological information and the like, but the method is not limited to the above factors, and other fault data can also be used as the above factors; the protection recorded image information may be recorded images in which a fault occurs for a certain time, for example, recorded images within k minutes and when a fault occurs.
In the implementation of the present disclosure, periodic acquisition of recorded wave images may be set, and a certain number of recorded wave images may be finally acquired. Taking 1 recorded image every 1 minute as an example, k+1 recorded images can be added up in k minutes, and the evolution process of the equipment fault in k minutes can be recorded by the k+1 recorded images. In the embodiment of the present disclosure, the acquisition manner is not limited, that is, the method is not limited to the manner of periodically acquiring the recorded image, and other manners, such as non-periodic acquisition and the manner of acquiring the recorded image based on the control instructions of other devices, are equally applicable to the embodiment of the present disclosure.
Illustratively, in the embodiments of the present disclosure, the textual information of the fault monitoring information may be obtained from a system, illustratively an intelligent regulation system, and the fault time information, such as the numerical information of the year, month, day, time, minute, second, etc. of the fault, may be extracted from the fault monitoring information. The acquired number of fault device recorded image information may be read from the intelligent regulation system, for example, k+1 recorded images as exemplified above may be acquired. After a certain number of fault equipment wave recording image information is obtained, a wave recording image capable of describing the fault process of the equipment is selected from the wave recording image information to serve as a wave recording image sample, the k value is obtained by analyzing a large number of historical fault wave recording image data, k can be set to be 5 in the embodiment of the invention, and the setting mode can be suitable for the requirement that the grid faults generally finish the evolution of the fault process within 5 minutes.
2. In the embodiment of the disclosure, the feature extraction and fusion of the fault monitoring information and the feature vector in the recorded wave image can be realized through a feature extraction and fusion module. The multi-source data feature extraction and fusion extraction module respectively extracts the feature vectors in the fault monitoring information and the recorded wave image, and fuses the extracted feature vectors of the two different data sources to form input variables of the equipment fault cause diagnosis model. The method comprises the following specific steps:
(1) The text type data is converted into a vector. In the embodiment of the disclosure, the fault monitoring information text can be converted into a vector through a pre-trained power professional language word vector model, each character in the fault monitoring information text can be converted into a 1-dimensional vector, wherein N elements are included, the 1-dimensional vector is formed by the values of the fault time of year, month, day, hour, minute and second according to a repeated mode, and the 1-dimensional vector and the text vector are spliced sequentially. Illustratively, for the time to failure (e.g., 2022-10-07 10:10:19) translate to [ A ] 1 … A 14 ]Vector, wherein there are 14 elements in the vector, converting each character in the faulty device description to [ B ] 1 … B N ]Vector, wherein there are N elements in the vector, then splice the two to [ A ] 1 … A 14 ； B 1 … B N ]. In the embodiment of the disclosure, only the conversion of two texts, namely, fault time and fault equipment description text information, is taken as an example for illustration, but the method is not limited to the two text information, is also not limited to specific vector dimensions, and can be correspondingly set according to the actual situation of a power grid.
(2) Extracting and processing image type dataIs characterized by (3). In the embodiment of the disclosure, feature selection and extraction can be performed on the image type variable of the wave-recording image based on a deep convolutional neural network method. The deep convolutional neural network is composed of a plurality of convolutional layers and pooling layers, so that the deep convolutional neural network can be used for carrying out rolling and pooling operations on the protected recorded image, extracting and dimension reduction on key image features in recorded image data are achieved, and flattening processing is carried out on the extracted data features. The input format of the data of the picture like the recorded image is set as (3, X, Y), wherein 3 represents the layer number of the deep convolutional neural network, and X, Y respectively represent the number of rows and the number of columns of the two-dimensional data. The output variable of the deep convolutional neural network is a one-dimensional eigenvector, the number of vector elements is N, the number of vector elements is consistent with the number of vector elements converted from the monitoring information text, and the output vector is [ C 1 …C N ]I.e. there are N elements inside the vector. The image characteristic result extracted by the deep convolutional neural network is normalized by adopting the following expression:
in the formula ：x sca a value normalized for the data;xthe method comprises the steps of extracting image characteristic data by using a deep convolutional neural network; max%x)、min(x) Respectively minimum value and maximum value of sample data, and min for image datax)、max(x) 0 and 255, respectively;u、vis the maximum and minimum of a given scaling range, and the general data range is set to be between (0, 1), which is taken in the embodiments of the present disclosureuCan be 0.9，vMay be 0.1. That is, as described above, in the embodiments of the present disclosure, the N-dimensional vector [ C ] is output 1 …C N ]Is the image characteristic output by the neural network after the normalization processing.
It should be noted that, the sequence is not limited to the processing of different types of data, the recorded wave image data can be processed first, and then the fault monitoring information can be processed; the fault monitoring information can be processed first, and then the recorded wave image data can be processed.
(3) The above-mentioned data of text type are converted into vectors and extracted characteristics of recorded wave image data are spliced together to form new characteristic vector-mixed coding vector which is used as input of power grid equipment fault cause diagnosis model, and the new characteristic vector, i.e. mixed coding vector of recorded wave image information and fault monitoring information is represented as [ A ] 1 …A 14 ；B 1 …B N ；C 1 …C N ]. From the new feature vector, it can be seen that the new feature vector fuses different types of information, i.e., monitoring information and recorded image information, at the same time.
3. In the embodiment of the disclosure, the cause probability of the fault device may be calculated by using the fault device cause analysis module, and by way of example, the power grid device fault cause diagnosis model construction module of the embodiment of the disclosure adopts a multi-sequence Long-Short-Term Memory network (M-LSTM) to perform encoding and decoding, and calculates a decoding result through a softmax layer in the multi-sequence Long-Term Memory network, thereby realizing the calculation of the cause probability of the fault device. The method comprises the following specific steps:
the multi-source data feature extraction and fusion extraction module obtains the new feature vector-hybrid coding vector, and in the embodiment of the disclosure, X is used i Hybrid code vector [ A ] representing the recorded image information and the fault-monitoring information 1 …A 14 ；B 1 …B N ；C 1 …C N ]Where i represents the number of samples in the hybrid encoded vector, exemplary hybrid encoded vector [ A ] 1 …A 14 ；B 1 …B N ；C 1 …C N ]There are 14 a samples, N B samples, N C samples, and a total of 14+n+m samples. It should be noted that in the embodiment of the present disclosure, the number of B samples and the number of C samples are not necessarily the same, and there may be M C samples, where n+.m. Forming the mixed coding vector and the fault category into a feature vector set
. In Y form j The fault cause category index number is represented, wherein j represents the number of the fault cause categories, and the fault cause categories can be mountain fire, thunder, strong wind, smoke dust, ice coating, ice flash, foreign matters and the like, and the index numbers of the fault cause categories can be-3, -2, -1, 0,1, 2 and 3 respectively.
Taking the mixed coding vector as an input of M-BiLSTM, and outputting the equipment failure cause as a diagnosis model by the M-BiLSTM, wherein the method comprises the following steps of: mountain fire, thunder and lightning, strong wind, smoke dust, ice coating, ice flash, foreign matters and the like.
The M-LSTM network improves the internal input of LSTM, the gate structure is not changed, the internal input is by introducing the actual value which can be acquired at the current moment
I.e. collected [ A ] 1 …A 14 ；B 1 …B N ；C 1 …C N ]And the corresponding actual failure cause, and utilizes the original predicted value
I.e. predicted [ A ] 1 …A 14 ；B 1 …B N ；C 1 …C N ]And the corresponding failure cause, the prediction error obtained by matrix calculation is input to the moment
In the corresponding door structure. I.e. M-LSTM network at time of day
The internal inputs at the time are:
wherein ,
predicting errors for fault reasons at the time t-1, namely the previous time; -1 is a negative identity matrix;
a predicted value vector of the data of the plurality of types at the time t-1;
the mixed coding vector is in the characteristic data after fusion at the time t-1;
is a matrix addition operator.
In M-LSTM networks, forget gate utilizes last time
Error in failure cause prediction
And the current time
Is of the fault feature vector of (a)
The forget control signal is adjusted. If the prediction error is too large, then it is explained that
For the current moment
Has no effective effect and therefore needs to forget this part of memory.
The input door utilizes the last moment
Error in failure cause prediction
And the current time
Is of the fault feature vector of (a)
The magnitude of the input control signal is adjusted. If prediction error
Too large, the current memory information needs to be reduced
Avoiding affecting the prediction of future time series.
The output door being likewise utilised
And
control and adjust the magnitude of the output control signal and determine the long-term memory
How much is output.
Input doorI t For controlling input of fault signature sequences in memory cells, forgetting gatesF t Determining how much information is kept in the memory cell at the previous time and outputting the information to the gateO t The output content is determined. The expression is as follows:
in the formula ：W i inputting door weights;h t-1 an implicit layer vector of the last time t-1;x t input data at the current t moment;b i an input gate bias term;
is a sigmoid function. Wherein:W f forgetting the door weight;b f a forget door bias term;W o outputting door weights;b o to output a gate bias term.
In an embodiment of the present disclosure,
for the memory cell at time t, the history memory content is stored, and after the reserved part of the past memory and the new content is determined, the cell is updated, and the expression is as follows:
in the formula ：W c is implicit state weight;b c biasing items for implicit states.
The embodiment of the disclosure is applied to a double-layer two-way long and short-term memory network (M-BiLSTM) through the improvement on the M-LSTM network, namely the M-BiLSTM is improved. Where the M-LSTM network is a single layer network, the exemplary implementation is a to B process, and the modified M-BiLSTM is a dual layer bi-directional, i.e., the first layer implements a to B process and the second layer implements B to a process.
Obtaining fault feature vectors of power grid equipment by adopting improved two-way long-short-term memory network (M-BiLSTM)
I.e. the equipment fault feature vector
Input into an improved two-way long and short term memory network. The M-BiLSTM is adopted to detect the fault characteristic vector and the fault cause (such as mountain fire, thunder, strong wind, smoke dust, ice coating, ice flash and foreign mattersEtc.) are encoded. In the embodiment of the disclosure, the hidden layer vector is obtained from the forward training of the equipment fault characteristic variable by utilizing the M-BiLSTM bidirectional networke t(p) Obtaining hidden layer vector from reverse training of equipment fault characteristic variablese t(q) By combininge t(p) Ande t(q) splicing the head and the tail to obtain an implicit layer vector for retaining fault characteristic variables of the bidirectional equipmente t By the hidden layer vectore t To enhance the device fault signature encoding capability. Wherein the parameters of M-LSTM and M-BiLSTM can be set as: the learning rate is 0.005, the batch size is 16, the optimization algorithm is Adam, the number of hidden layers is 2, the number of hidden layers is 256, the random inactivation rate of neurons is 0.5, and the maximum training iteration number is 200.
And calculating the identification relation between the monitoring information and the equipment protection recorded image fusion characteristic information and the cause of the fault equipment through softmax, wherein the expression is as follows:
wherein: e is a natural constant, and is a natural constant,s i for the equipment failure reason label probability, K is equipment failure category number, and the failure category comprises: mountain fire, thunder and lightning, strong wind, smoke dust, ice coating, ice flash, foreign matters and the like.
In the embodiment of the disclosure, the online diagnosis module for the power grid equipment fault cause can sense various types of data such as texts and images of fault monitoring information, wave recording image information and the like in real time, process the data and fuse the data to form fused feature vectors, and map the fused feature vectors into an equipment fault cause diagnosis model to obtain corresponding fault cause diagnosis results.
On the basis of the disclosure, the embodiment of the disclosure also provides a power grid fault diagnosis device, and the power grid fault diagnosis device of the disclosure example shown in fig. 3 comprises an acquisition unit, a fusion unit and an analysis unit. The device comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for preprocessing various types of data in a power grid and acquiring characteristic data of the various types of data; the fusion unit is used for fusing the characteristic data of the multiple types of data to form fused characteristic data; and the analysis unit is used for analyzing the power grid equipment faults according to the fused characteristic data. Further processing steps performed by the acquisition unit, the fusion unit and the analysis unit are as described above and are not described in detail herein.
Based on the disclosure, correspondingly, the invention further provides electronic equipment. As shown in fig. 4, an electronic device of an embodiment of the present disclosure includes at least one processor and a memory electrically connected to the processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described above.
In the embodiment of the disclosure, the trained power grid equipment fault cause diagnosis model can be deployed in a safety I area of the intelligent regulation system, and factors such as fault equipment description, fault time, power flow mutation equipment description, power grid operation mode information description, meteorological information and the like and protection wave recording image information are acquired in real time. The information is subjected to a multi-source data preprocessing module and a multi-source data characteristic extraction and fusion extraction module to obtain a multi-source data fusion characteristic vector, the multi-source data fusion characteristic vector is input into a power grid equipment fault cause diagnosis model, and equipment fault causes can be mapped out through the model rapidly, and the method comprises the following steps: mountain fire, thunder and lightning, strong wind, smoke dust, ice coating, ice flash, foreign matters and the like.
It should be noted that, each module in the embodiments of the present disclosure is not necessarily an actual hardware component, and may be each processing subroutine in a computer program, each data processing subunit in an electronic circuit, and each data processing portion in a processor.
Although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (11)
1. A power grid fault diagnosis method based on multi-class information fusion identification, the method comprising:
preprocessing various types of data in a power grid to obtain characteristic data of the various types of data;
fusing the characteristic data of the multiple types of data comprising the fault monitoring information and the wave recording image information to form fused characteristic data; the fused characteristic data comprises a mixed coding vector of the multiple types of data and fault class marks, wherein the mixed coding vector comprises fault monitoring information and wave recording image information;
the method comprises the steps of acquiring and fusing fault monitoring information and wave recording image information, and specifically comprises the following steps: converting fault monitoring information text into text vectors by using a pre-trained power professional language word vector model, splicing fault time vectors and the text vectors, performing feature selection and extraction on the characteristics of the recorded image data based on a deep convolutional neural network, and splicing the text vectors spliced with the fault time vectors and the recorded image data characteristics to obtain mixed coding vectors;
analyzing the power grid equipment faults according to the fused characteristic data;
after the fused characteristic data are formed, carrying out logic operation on the fused characteristic data and the predicted characteristic data to obtain a fault cause prediction error at the previous moment; wherein, the fault cause prediction error of the previous moment is obtained by the following way:
in the formula ,
for fault reasons at time t-1, i.e. the previous timeError measurement, -1 is a negative identity matrix,>
for the vector of predictors of said plurality of types of data at time t-1 +.>
For the mixed coding vector in the feature data after fusion at time t-1 +.>
Is a matrix addition operator;
determining an input gate signal, a forget gate signal and an output gate signal in a gate structure in an M-LSTM network based on the fault cause prediction error at the previous moment and the fault feature vector at the current moment;
and finally, determining the fault of the power grid equipment.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises,
the predicted values of the plurality of types of data are obtained through an LSTM algorithm.
3. The method of claim 1, wherein the incoming gating signal, the forget gating signal, and the outgoing gating signal in the gating structure in the M-LSTM network are determined by:
wherein ,
I t for inputting gate signals for controlling the input of fault signature sequences in memory cells,F t for forgetting the gate signal, for determining how much information was kept at the previous time in the current memory cell,O t determining output content for the output gate;
W i inputting door weights; x is x t At time tInputting data; b i An input gate bias term;
W f forgetting the door weight; b f A forget door bias term; w (W) o Outputting door weights; b o To output a gate bias term.
4. A method according to claim 3, characterized in that the memory cells at time t are updated by
：
wherein ,W c is implicit state weight;b c biasing items for implicit states.
5. The method according to any one of claims 1 to 4, wherein,
taking the mixed coding vector in the fused characteristic data as the input of M-BiLSTM;
outputting a fault reason category number by using the M-BiLSTM;
and determining the power grid fault reason according to the fault reason category number.
6. A power grid fault diagnosis device based on multi-class information fusion identification, the device comprising:
the device comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit is used for preprocessing various types of data in a power grid and acquiring characteristic data of the various types of data;
the fusion unit is used for fusing the characteristic data of the multiple types of data comprising the fault monitoring information and the wave recording image information to form fused characteristic data; the fused characteristic data comprises a mixed coding vector of the multiple types of data and fault class marks, wherein the mixed coding vector comprises fault monitoring information and wave recording image information; the method comprises the steps of acquiring and fusing fault monitoring information and wave recording image information, and specifically comprises the following steps: converting fault monitoring information text into text vectors by using a pre-trained power professional language word vector model, splicing fault time vectors and the text vectors, performing feature selection and extraction on the characteristics of the recorded image data based on a deep convolutional neural network, and splicing the text vectors spliced with the fault time vectors and the recorded image data characteristics to obtain mixed coding vectors;
the analysis unit is used for analyzing the power grid equipment faults according to the fused characteristic data; after the fused characteristic data are formed, the analysis unit carries out logic operation on the fused characteristic data and the predicted characteristic data to obtain a fault cause prediction error at the previous moment; wherein, the fault cause prediction error of the previous moment is obtained by the following way:
，/>
for the prediction error of the cause of the fault at time t-1, i.e. the previous time, -1 is a negative identity matrix,/>
For the vector of predictors of said plurality of types of data at time t-1 +.>
For the mixed coding vector in the feature data after fusion at time t-1 +.>
Is a matrix addition operator; determining the input gate signal and forgetting in the gate structure in the M-LSTM network based on the fault cause prediction error of the previous moment and the fault feature vector of the current momentGate signals and output gate signals; and finally, determining the fault of the power grid equipment.
7. The apparatus of claim 6, wherein the device comprises a plurality of sensors,
the predicted values of the plurality of types of data are obtained through an LSTM algorithm.
8. The apparatus of claim 6, wherein the ingress gating signal, the forget gating signal, and the egress gating signal in the gate structure in the M-LSTM network are determined by:
wherein ,
I t for inputting gate signals for controlling the input of fault signature sequences in memory cells,F t for forgetting the gate signal, for determining how much information was kept at the previous time in the current memory cell,O t determining output content for the output gate;
W i inputting door weights; x is x t Input data at the time t; b i An input gate bias term;
W f forgetting the door weight; b f A forget door bias term; w (W) o Outputting door weights; b o To output a gate bias term.
9. The device of claim 8, wherein the memory cells at time t are updated by
：
wherein ,W c is implicit state weight;b c biasing items for implicit states.
10. The device according to any one of claims 6 to 9, wherein,
taking the mixed coding vector in the fused characteristic data as the input of M-BiLSTM;
outputting a fault reason category number by using the M-BiLSTM;
and determining the power grid fault reason according to the fault reason category number.
11. An electronic device comprising at least one processor and a memory electrically connected to the processor, wherein the memory stores instructions for execution by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211512825.2A CN115545355B (en) | 2022-11-30 | 2022-11-30 | Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211512825.2A CN115545355B (en) | 2022-11-30 | 2022-11-30 | Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115545355A CN115545355A (en) | 2022-12-30 |
| CN115545355B true CN115545355B (en) | 2023-05-02 |
Family
ID=84722207
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211512825.2A Active CN115545355B (en) | 2022-11-30 | 2022-11-30 | Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115545355B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112712205A (en) * | 2020-12-29 | 2021-04-27 | 南京后生远达科技有限公司 | Power distribution network fault prevention method based on long-term and short-term memory neural network |
| CN113325266A (en) * | 2021-05-19 | 2021-08-31 | 国网辽宁省电力有限公司 | Power grid fault diagnosis method and system based on fuzzy integral multi-source information fusion |
| CN114498933A (en) * | 2022-02-09 | 2022-05-13 | 航天科工深圳（集团）有限公司 | Load identification and multi-sensor fusion sensing method and device |
| CN114707754A (en) * | 2022-04-25 | 2022-07-05 | 广西电网有限责任公司 | Intelligent ammeter fault prediction method and system based on BiLSTM-CNN model |
| CN115270881A (en) * | 2022-08-01 | 2022-11-01 | 国网宁夏电力有限公司固原供电公司 | Line fault recognition method based on multi-mode information fusion knowledge graph |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108226711A (en) * | 2018-01-05 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of electric network failure diagnosis method and device based on information fusion |
| CN110232395B (en) * | 2019-03-01 | 2023-01-03 | 国网河南省电力公司电力科学研究院 | Power system fault diagnosis method based on fault Chinese text |
| CN112748331A (en) * | 2020-12-24 | 2021-05-04 | 国网江苏省电力有限公司电力科学研究院 | Circuit breaker mechanical fault identification method and device based on DS evidence fusion |
| CN113159345A (en) * | 2021-05-08 | 2021-07-23 | 中国电力科学研究院有限公司 | Power grid fault identification method and system based on fusion neural network model |
| CN114528897A (en) * | 2021-07-27 | 2022-05-24 | 河北工业大学 | Equipment fault diagnosis method based on knowledge and data fusion drive |
-
2022
- 2022-11-30 CN CN202211512825.2A patent/CN115545355B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112712205A (en) * | 2020-12-29 | 2021-04-27 | 南京后生远达科技有限公司 | Power distribution network fault prevention method based on long-term and short-term memory neural network |
| CN113325266A (en) * | 2021-05-19 | 2021-08-31 | 国网辽宁省电力有限公司 | Power grid fault diagnosis method and system based on fuzzy integral multi-source information fusion |
| CN114498933A (en) * | 2022-02-09 | 2022-05-13 | 航天科工深圳（集团）有限公司 | Load identification and multi-sensor fusion sensing method and device |
| CN114707754A (en) * | 2022-04-25 | 2022-07-05 | 广西电网有限责任公司 | Intelligent ammeter fault prediction method and system based on BiLSTM-CNN model |
| CN115270881A (en) * | 2022-08-01 | 2022-11-01 | 国网宁夏电力有限公司固原供电公司 | Line fault recognition method based on multi-mode information fusion knowledge graph |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115545355A (en) | 2022-12-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111914873A (en) | Two-stage cloud server unsupervised anomaly prediction method | |
| CN111428789A (en) | Network traffic anomaly detection method based on deep learning | |
| CN111310672A (en) | Video emotion recognition method, device and medium based on time sequence multi-model fusion modeling | |
| CN111353153A (en) | GEP-CNN-based power grid malicious data injection detection method | |
| CN115220133B (en) | Rainfall prediction method, device and equipment for multiple meteorological elements and storage medium | |
| CN115496384A (en) | Monitoring management method and device for industrial equipment and computer equipment | |
| CN117154263A (en) | Lithium battery cascade utilization charging and discharging system and control method | |
| CN114662791A (en) | Long time sequence pm2.5 prediction method and system based on space-time attention | |
| CN115221233A (en) | Transformer substation multi-class live detection data anomaly detection method based on deep learning | |
| KR102540027B1 (en) | Power control system and method characterized in that possible to predict power demand and detect abnormalities | |
| CN113758652B (en) | Oil leakage detection method and device for converter transformer, computer equipment and storage medium | |
| CN115545355B (en) | Power grid fault diagnosis method, device and equipment based on multi-class information fusion recognition | |
| CN117009785A (en) | Security monitoring method, device, server and system | |
| CN117113202A (en) | Power loop energy consumption detection method and equipment based on joint error stacking model | |
| CN115409100A (en) | Method, device, medium and equipment for generating power accident emergency scheme | |
| CN114819350A (en) | RUL prediction method under sensor fault condition based on multiple tasks | |
| CN112232557B (en) | Short-term prediction method for health degree of switch machine based on long-short-term memory network | |
| CN114595952A (en) | Electricity stealing behavior detection method based on attention network improved convolutional neural network | |
| CN109800923B (en) | Short-term power combination prediction method for distributed wind power generation | |
| CN110837932A (en) | Thermal power prediction method of solar heat collection system based on DBN-GA model | |
| CN117805607B (en) | DC level difference matching test method for power plant DC system | |
| CN118069717B (en) | Time sequence data characteristic prediction method based on cyclic neural network | |
| Zheng et al. | Wind Electricity Power Prediction Based on CNN-LSTM Network Model | |
| JP7475549B2 (en) | Learning device, prediction device, prediction system, learning method, prediction method, and prediction program | |
| CN117667495B (en) | Association rule and deep learning integrated application system fault prediction method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |